diff --git a/package.json b/package.json
index edd830f..3045b0b 100644
--- a/package.json
+++ b/package.json
@@ -40,7 +40,12 @@
 			  "ollama-coder.max-tokens-predicted": {
 				"type": "integer",
 				"default": 500,
-				"description": "The system message to use for code completions. Type DEFAULT for Makefile."
+				"description": "The maximum number of tokens generated by the model."
+			  },
+			  "ollama-coder.prompt-window-size": {
+				"type": "integer",
+				"default": 2000,
+				"description": "The size of the prompt in characters. NOT tokens, so can be set about 1.5-2x the max tokens of the model (varies)."
 			  }
 		  }
 		}
diff --git a/src/extension.ts b/src/extension.ts
index ba75afc..3a64a47 100644
--- a/src/extension.ts
+++ b/src/extension.ts
@@ -1,22 +1,24 @@
 // Significant help from GPT4
 
 import * as vscode from "vscode";
-import axios from "axios";
+import axios, { AxiosResponse } from "axios";
 
-const apiEndpoint: string = vscode.workspace.getConfiguration("ollama-coder").get("apiEndpoint") || "http://localhost:11434/api/generate";
-const apiModel: string = vscode.workspace.getConfiguration("ollama-coder").get("model") || "deepseek-coder";
-let apiSystemMessage: string | undefined = vscode.workspace.getConfiguration("ollama-coder").get("system-message");
+const VSConfig = vscode.workspace.getConfiguration("ollama-coder");
+const apiEndpoint: string = VSConfig.get("apiEndpoint") || "http://localhost:11434/api/generate";
+const apiModel: string = VSConfig.get("model") || "deepseek-coder";
+let apiSystemMessage: string | undefined = VSConfig.get("system-message");
 if (apiSystemMessage == "DEFAULT") apiSystemMessage = undefined;
-const numPredict: number = vscode.workspace.getConfiguration("ollama-coder").get("max-tokens-predicted") || 500;
+const numPredict: number = VSConfig.get("max-tokens-predicted") || 500;
+const promptWindowSize: number = VSConfig.get("prompt-window-size") || 2000;
 
 // This method is called when your extension is activated
 function activate(context: vscode.ExtensionContext) {
-	console.log("Ollama Coder is Active");
 	// Register a completion provider for JavaScript files
-	const provider = vscode.languages.registerCompletionItemProvider("javascript", {
-		async provideCompletionItems(document, position) {
+	const provider = vscode.languages.registerCompletionItemProvider("*", {
+		async provideCompletionItems(document, position, cancellationToken) {
 			// Get the current prompt
-			const prompt = document.getText(new vscode.Range(document.lineAt(0).range.start, position));
+			let prompt = document.getText(new vscode.Range(document.lineAt(0).range.start, position));
+			prompt = prompt.substring(Math.max(0, prompt.length - promptWindowSize), prompt.length);
 			// Check if the prompt is not empty and ends with a dot
 			if (prompt) {
 				// Create a completion item
@@ -29,7 +31,7 @@ function activate(context: vscode.ExtensionContext) {
 				item.command = {
 					command: 'ollama-coder.autocomplete',
 					title: 'Ollama',
-					arguments: [document, position, prompt]
+					arguments: [document, position, prompt, cancellationToken]
 				};
 				// Return the completion item
 				return [item];
@@ -45,62 +47,91 @@ function activate(context: vscode.ExtensionContext) {
 	// Register a command for getting a completion from Ollama
 	const disposable = vscode.commands.registerCommand(
 		"ollama-coder.autocomplete",
-		async function (document, position, prompt) {
+		async function (document: vscode.TextDocument, position: vscode.Position, prompt: string, cancellationToken: vscode.CancellationToken) {
 			// Show a progress message
 			vscode.window.withProgress(
 				{
 					location: vscode.ProgressLocation.Notification,
 					title: "Getting a completion from Ollama...",
+					cancellable: true,
 				},
-				async (progress, token) => {
+				async (progress, progressCancellationToken) => {
 					try {
 						// Make a request to the ollama.ai REST API
 						const response = await axios.post(apiEndpoint, {
 							model: apiModel, // Change this to the model you want to use
 							prompt: prompt,
-							stream: false,
+							stream: true,
 							system: apiSystemMessage,
 							options: {
 								num_predict: numPredict
-							}
+							},
+						}, {
+							cancelToken: new axios.CancelToken((c) => {
+								const cancelPost = function () {
+									c("Autocompletion request terminated");
+								};
+								cancellationToken.onCancellationRequested(cancelPost);
+								progressCancellationToken.onCancellationRequested(cancelPost);
+								vscode.workspace.onDidCloseTextDocument(cancelPost);
+							}),
+							responseType: 'stream'
 						}
 						);
-						// Get the completion from the response
-						const completion = response.data.response;
-						// Check if the completion is not empty
-						if (completion) {
-							// Insert the completion into the document
+
+						//tracker
+						let currentPosition = position;
+
+						response.data.on('data', async (d: Uint8Array) => {
+							// Get a completion from the response
+							const completion: string = JSON.parse(d.toString()).response;
+
+							//complete edit for token
 							const edit = new vscode.WorkspaceEdit();
 							const range = new vscode.Range(
-								position.line,
-								position.character,
-								position.line,
-								position.character
+								currentPosition.line,
+								currentPosition.character,
+								currentPosition.line,
+								currentPosition.character
 							);
 							edit.replace(document.uri, range, completion);
 							await vscode.workspace.applyEdit(edit);
+
 							// Move the cursor to the end of the completion
+							const completionLines = completion.split("\n");
 							const newPosition = position.with(
-								position.line,
-								position.character + completion.length
+								currentPosition.line + completionLines.length,
+								(completionLines.length > 0 ? 0 : currentPosition.character) + completionLines[completionLines.length - 1].length
 							);
 							const newSelection = new vscode.Selection(
 								newPosition,
 								newPosition
 							);
+							currentPosition = newPosition;
+
+							// completion bar
+							progress.report({ increment: 1 / (numPredict/100) });
+
+							// move cursor
 							const editor = vscode.window.activeTextEditor;
 							if (editor) editor.selection = newSelection;
-						} else {
-							// Show a warning message
-							vscode.window.showWarningMessage("Ollama could not generate a completion for this prompt");
-							console.log("Ollama could not generate a completion for this prompt");
-						}
+						});
+
+						// Keep cancel window available
+						const finished = new Promise((resolve) => {
+							response.data.on('end', () => {
+								progress.report({ message: "Ollama completion finished." });
+								resolve(true);
+							});
+						});
+
+						await finished;
+
 					} catch (err: any) {
 						// Show an error message
 						vscode.window.showErrorMessage(
 							"Ollama encountered an error: " + err.message
 						);
-						console.log("Ollama encountered an error: " + err.message);
 					}
 				}
 			);